DigitRecog

Digit recognition neural network using the MNIST dataset, with support for batch and online descent. This code IS documented, so read away! The math involved, though is not, so you may want to familiarize yourself with it before trying to read the docs. No bugs have been observed. This model has up to 94-95% accuracy, probably due to its architecture. I am aware the NN class is extremely verbose, but it was coded this way as a means of making the obtuse math involved more explicit and easy to understand. I may come back to redo it later, but as it should have little to no effect on speed it is a low priority.